I don't think SQL gets replaced until we move to a different database paradigm. For example, I could imagine a function database model [0] getting paired with a "grammar of graphics" type of metadata for the records to create much more concise query/aggregate declarations, especially where time series are concerned.
SQL, for the most part, has already been replaced as a language used by people with alternatives like query builders, ORMs, etc. Beyond the occasional ad-hoc query, SQL simply isn't suitable for the tasks we require of modern relational databases, lacking features like composition that today's applications need.
It remains as a compiler target for those tools, but I am not sure that is the level of abstraction in question.
Comments
I don't think SQL gets replaced until we move to a different database paradigm. For example, I could imagine a function database model [0] getting paired with a "grammar of graphics" type of metadata for the records to create much more concise query/aggregate declarations, especially where time series are concerned.
[0] https://en.wikipedia.org/wiki/Functional_database_model
SQL, for the most part, has already been replaced as a language used by people with alternatives like query builders, ORMs, etc. Beyond the occasional ad-hoc query, SQL simply isn't suitable for the tasks we require of modern relational databases, lacking features like composition that today's applications need.
It remains as a compiler target for those tools, but I am not sure that is the level of abstraction in question.